I’m no electronics expert, but I think there are two main angles on this.
For a given amp draw at the LED, the battery needs to supply a higher current due to loses from the driver/wiring. Good li-ion can supply high current, but when batteries supply high current their voltage drops. So while they may have a resting voltage of ‘x’, when under load their voltage becomes ‘x - y’.
For example, most protected li-ion cells will trip the protection circuit at 2.75v. But I’ve had batteries that have a resting voltage of 3.5v trip this circuit, what happens is, when under load it dips below 2.75v so trips the protection.
To maintain output from the LED, you need voltage as well as current. An example here would be, an MT-G2 will supply lots of lumens at 3amps. A good single 18650 can easily supply 3amps. But the vf of the MT-G2 is a higher voltage than that of a single li-ion. So despite being able to supply the amps, there is insufficient voltage to power the LED.
Using multiple Li-ion in series means you have a higher input voltage than just using one battery. So for LED’s like XP-G and XM-L’s, your input voltage is always sufficient (ie higher) than needed. This means under load, even when the batteries voltage sag, you’ll still be ok on the volts front. Multiple cells is also kinder in terms of amp draw. If your driver is pulling 3amps, then with two batteries it only needs to pull 1.5amp from each.
Overall this may not affect total PEAK output of a torch, but it can affect regulation, depending on the driver type and circuit used.
And driver type is the other key point here. There are different types of driver, I’m not clever enough to explain them properly. But as I understand it, in simply terms. A driver can use more mAh to boost the voltage it provides to the LED. e.g. many AA lights. If you simply connected the LED directly to a 1.2v NiMH eneloop, it wouldn’t be all that bright. But a boost driver can sacrifice some mAh/capacity too boost the voltage it supplies to the LED. Thus more lumens.
However this can be taxing on a battery and will lower overall run time.
Then you have buck drivers where you have more voltage than you need, here the driver must reduce the voltage supplied to the LED. Sometimes less efficient, but offers greater stable output and is kinder to batteries. Excessively high voltage can mean large efficiency loses and heat.
Most drivers can do one or the other of these, hence why most are voltage limited. But some can do both, which are the ones with the widest voltage ranges. The trouble is, to make them do both, they often end up being less efficient at both ends of the scale.
A nice example of this is too take a look at the MagLed D cell lights (the 2nd gen ones). The 2D and 3D both produce almost identical lumen output.
2D = 134 lumens
3D = 131 lumens.
People often wonder why the 3D has a lower output. And this is down to the driver and I suspect how they are configured. Mag tried to get them similar, but it’s probably difficult to get them performing the same.
The 2D uses 2x1.5v, so 3 volt battery power. This is lower than the LED vf, so this is a boost driver. It uses mAh to raise the voltage supplied to the LED.
The 3D is 3x1.5v, so 4.5 volt battery power. This is higher than the LED vf, although not by much. This is probably a buck or maybe linear type of driver.
Why does this matter?
Well, considering they have similar output and the 3D only uses one more battery. You’d think they would have similar runtimes too. Maybe the 3D should last half as long again than what the 2D does.
But nope.
2D = 8 hour
3D = 79 hour!!!
That’s a huge difference. And what it shows is, the 2D actually has to work the cells fairly hard compared to the 3D, just to match the output.
How does this relate to your question?
Well in Mag’s case they chose to make the 3D offer the same amount of light as the 2D. But higher input voltage could have meant they could have driven the 3D batteries as hard as the 2D ones. If they’d done this the batteries would be no worse off, but you’d have got more lumen output from them and still matched the 8 hour runtime.
There are many options here. Tine foil packed around them. Copper tape. Cut up Coke can’s or remove the large spring and use a bit of copper pipe cut to the right length.